sign in sign up
<<
Home , Aethotaxis mitopteryx, Artedidraco, Bald notothen, Bathydraco, Channichthyidae, Gobionotothen acuta

0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier.com/locate/ympev

Phylogenetic analysis of Antarctic notothenioids illuminates the utility of 7 RADseq for resolving Cenozoic adaptive radiations

Thomas J. Neara,b,⁎, Daniel J. MacGuigana, Elyse Parkera, Carl D. Struthersc, Christopher D. Jonesd, Alex Dornburge a Department of Ecology & Evolutionary Biology, Yale University, P.O. Box 208106, New Haven, CT 06520, USA b Peabody Museum of Natural History, Yale University, New Haven, CT 06520, USA c Museum of New Zealand Te Papa Tongarewa, Wellington, New Zealand d Antarctic Ecosystem Research Division, NOAA Southwest Fisheries Science Center, La Jolla, CA 92037, USA e North Carolina Museum of Natural Sciences, Raleigh, NC 27601, USA

ARTICLE INFO ABSTRACT

Keywords: Notothenioids are a clade of ∼120 of marine fishes distributed in extreme southern hemisphere tem- perate near-shore habitats and in the surrounding . Over the past 25 years, molecular Phylogenomics and morphological approaches have redefined hypotheses of relationships among notothenioid lineages as well Phylogenetic informativeness as their relationships among major lineages of percomorph teleosts. These phylogenies provide a basis for in- Signal noise vestigation of mechanisms of evolutionary diversification within the clade and have enhanced our understanding Icefishes of the notothenioid adaptive radiation. Despite extensive efforts, there remain several questions concerning the phylogeny of notothenioids. In this study, we deploy DNA sequences of ∼100,000 loci obtained using RADseq to investigate the phylogenetic relationships of notothenioids and to assess the utility of RADseq loci for lineages that exhibit divergence times ranging from the Paleogene to the Quaternary. The notothenioid phylogenies inferred from the RADseq loci provide unparalleled resolution and node support for several long-standing problems including, (1) relationships among species of , (2) resolution of Indonotothenia cyano- brancha as the sister lineage of Trematomus, (3) the deep paraphyly of , (4) the paraphyly of s.l., (5) paraphyly of , and 6) the monophyly of the . Assessment of site rates demonstrates that RADseq loci are similar to mtDNA protein coding genes and exhibit peak phylo- genetic informativeness at the time interval during which the major Antarctic notothenioid lineages originated and diversified. In addition to providing a well-resolved phylogenetic hypothesis for notothenioids, our analyses quantify the predicted utility of RADseq loci for Cenozoic phylogenetic inferences.

1. Introduction economic importance for international fisheries interests (Constable et al., 2000; Abrams, 2013). Despite a long history of research, a well- Antarctic notothenioids (Cryonotothenioidea) are one of the most resolved species-level phylogeny of notothenioids is not available and well studied groups of marine fishes and one of few examples of marine several key phylogenetic questions remain unanswered. teleost adaptive radiation (Eastman, 1993; Clarke and Johnston, 1996; Over the past quarter-century efforts to investigate the phyloge- Ingram and Mahler, 2011; Matschiner et al., 2011; Near et al., 2012; netics of notothenioids have resulted in important discoveries that Colombo et al., 2015; Dornburg et al., 2017a). In addition to exhibiting dramatically altered subsequent taxonomic classifications. For example, an interesting evolutionary history, notothenioids are vital to Antarctic early morphological and molecular inferred phylogenies resolve marine ecosystems, as they comprise a substantial component of the Bovichtidae, historically delimited to include , , and biomass, abundance, and species diversity of near shore fishes Pseudaphritis(Eastman, 1993; Nelson, 1994), as paraphyletic and Ele- (Eastman, 1993, 2005). Correspondingly, these species are critical in ginops maclovinus as the sister lineage of Cryonotothenioidea instead of linking lower level consumers and higher level predators in the Ant- being closely related to the nototheniid lineage (Balushkin, arctic marine food web (La Mesa et al., 2004), including species of high 1992; Lecointre et al., 1997; Bargelloni and Lecointre, 1998). The most

⁎ Corresponding author at: Department of Ecology & Evolutionary Biology, Yale University, P.O. Box 208106, New Haven, CT 06520, USA. E-mail address: [email protected] (T.J. Near). https://doi.org/10.1016/j.ympev.2018.09.001 Received 10 April 2018; Received in revised form 31 August 2018; Accepted 1 September 2018 $YDLODEOHRQOLQH6HSWHPEHU ‹(OVHYLHU,QF$OOULJKWVUHVHUYHG T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ² recent phylogenetic analyses of notothenioids use DNA sequences 2007; Townsend et al., 2012). Our results provide a strongly supported sampled from multiple mitochondrial and nuclear genes with a taxon phylogenetic hypothesis of notothenioid species-level relationships, the sampling that includes most of the recognized species in the clade (Near ability to reject several previous phylogenetic hypotheses and refine our et al., 2012; Dornburg et al., 2017a). While these studies provide im- perspective on the predicted utility and limits of RADseq data. portant insights into the relationships of notothenioids and serve as the basis for comparative analyses investigating the history and mechan- 2. Methods isms of notothenioid diversification (Near et al., 2012; Dornburg et al., 2017a), there are several unresolved issues in the phylogenetics of 2.1. Taxon sampling, sequencing, and data preparation notothenioids: the lack of phylogenetic resolution among the ∼14 species of the rapidly diversifying Trematomus (Kuhn and Near, 2009; The taxon sampling included 80 notothenioid species Janko et al., 2011; Lautredou et al., 2012; Near et al., 2012); the re- (Supplementary Table 1), which encompasses all of the recognized solution of the neutrally buoyant Pleuragramma antarctica among the taxonomic families and is very similar to previous phylogenetic studies major lineages of Cryonotothenioidea (Near and Cheng, 2008; Dettaï using Sanger sequenced legacy markers (Near et al., 2012). For ex- et al., 2012; Near et al., 2012); and the lack of support for monophyly of ample, the species included in Near et al. (2012) that are not included the Antarctic Dragonfishes (Bathydraconidae) in molecular analyses in this study are acuta, G. marionensis, (Bargelloni et al., 2000; Derome et al., 2002; Dettaï et al., 2012; Near angustata, scotiae, , and ant- et al., 2012). These remaining challenges to the resolution of notothe- arcticus. Species included in this study that are not in Near et al. (2012) nioid phylogeny inhibit the investigation of important questions, such are cf. nudifrons, Pogonophryne fusca, and Bathydraco as the origin of neutral buoyancy (Near et al., 2007, 2012), species joannae. We are confident that the minor differences in taxon sampling relationships within rapidly diversifying lineages (e.g., Trematomus & do not have a strong effect on the phylogenetic inferences. ; Lecointre et al., 2011; Lautredou et al., 2012), and DNA was isolated from tissue samples using the Qiagen DNeasy patterns of evolution in Bathydraconidae that led to the tissue extraction protocols (DNeasy, Qiagen, Valencia, CA). Extractions loss of this protein in the Crocodile Icefishes () were gel-quantified in agarose using New England Biolabs 100 bp (Bargelloni et al., 1998; Near et al., 2006; Lau et al., 2012). ladder (NEB, Ipswich, MA) to ensure successful extractions and DNA Next-generation sequencing through reduced representation concentrations were determined using a Qubit v. 3.0 fluorometer methods such as restriction site associated DNA sequencing (RADseq) (ThermoFisher Scientific, Philadelphia, PA). All DNA extractions were hold the promise of resolving species-level relationships in notothe- standardized to contain between 17 and 23 ng DNA/µL. nioids. By sequencing DNA flanking restriction sites, RADseq captures The RADseq protocol was not optimized for notothenioids, but ra- thousands of single nucleotide polymorphisms (SNPs) across any target ther is a single enzyme protocol that has been used in lineages of and have been used to resolve difficult phylogenetic problems flowering plants (Eaton et al., 2017). Floragenex Inc. (Portland, OR) in lineages spanning beetles (Cruaud et al., 2014), plants (Massatti prepared the RADseq libraries using the SbfI restriction enzyme, a six et al., 2016; Wang et al., 2017), corals (Herrera and Shank, 2016), and base pair cutter (5′-CCTGCA-3′) and sample-specific barcodes. The Lake Victoria cichlids (Wagner et al., 2013). Empirical assessments of samples were combined into two 95-sample multiplexed libraries. RADseq data suggest cost-effective phylogenetic utility across temporal Floragenex created two replicates of each library in order to minimize scales spanning recent divergences to those dating back tens of millions the influence of PCR duplication bias and other technical errors. We of years (Eaton et al., 2017). However, these assessments are based on sequenced each library twice on an Illumina HiSeq 2000 using single- post-hoc measure of topological support such as bootstrap values that end 100 base pair sequencing at the University of Oregon GC3F facility can be positively misled by noise in the dataset, or phylogenetic in- (https://gc3f.uoregon.edu/). We used pyrad v.3.0.61 to assemble and formativeness profiles (Townsend, 2007), which make no explicit pre- align the RADseq datasets (Eaton, 2014). Individual reads that con- diction of phylogenetic noise or tree topology (Townsend, 2007; Collins tained more than four sites with Phred scores < 20 were excluded. and Hrbek, 2018). Reads were clustered using vsearch and an 88% similarity threshold, It is important to consider that the length of a given internode can allowing for approximately 11 base differences between reads within a fundamentally alter expectations of phylogenetic utility (Whitfield and cluster. Analyses did not include clusters with a sequencing depth of Lockhart, 2007; Steel and Leuenberger, 2017; Dornburg et al., 2018). A less than six reads. We also discarded consensus sequences for each relationship exists between the timescale of a phylogenetic question cluster if they contained more than five heterozygous or ambiguous

(T), the time between branching events (t0), and the rate of character bases. Consensus sequences were clustered as homologous loci across evolution, which allows direct quantification of the predictive utility of samples with an 88% similarity threshold, excluding loci that were a given marker (Townsend et al., 2012; Su et al., 2014). A rapidly shared by fewer than four of the sampled specimens. Accession num- evolving marker will provide phylogenetic information for deep di- bers for the raw sequence reads deposited in the NCBI BioSample da- vergences when waiting times between branching events are long, tabase are given in Supplementary Table 1. while the same marker has a higher probability of being positively misleading at a similar timescale when waiting times between 2.2. Phylogenomic inference and testing alternative phylogenetic hypotheses branching events are reduced (Townsend et al., 2012). Although pre- vious research has demonstrated that RADseq successfully resolves The RADseq alignment was analyzed using IQ-TREE to determine phylogenetic relationships dating ∼50 million years (Eaton et al., the optimal molecular evolutionary model and infer a maximum like- 2017; Collins and Hrbek, 2018), no empirical assessments have ana- lihood phylogeny of notothenioids (Nguyen et al., 2015; lyzed the conditions under which RADseq remains cost-effective versus Kalyaanamoorthy et al., 2017). Node support was assessed using an potentially misleading for phylogenetic inference. ultrafast bootstrap analysis with 1000 replicates (Hoang et al., 2018). Here we assemble a RADseq dataset for the majority of living no- Based on previous analyses (e.g., Near et al., 2012), we rooted our tothenioid species, providing the first detailed phylogenomic in- phylogenetic inferences using the species of Bovichtidae sampled in this vestigations of this radiation. To evaluate the impact of internode study. length on the predictive utility of sequenced RADseq loci, we assess In order to accommodate incomplete lineage sorting, we used the patterns of phylogenetic information content across tens of thousands species tree approach tetrad v.0.7.19, an implementation of of RADseq loci using a combination of phylogenetic informativeness SVDquartets in the software package iPyrad (Chifman and Kubatko, (PI) that complement those presented in Collins and Hrbek (2018), and 2014; http://github.com/dereneaton/ipyrad). We ran tetrad using one approaches to quantifying phylogenetic signal and noise (Townsend, randomly selected SNP per locus, for 102,232 SNPs. We inferred all

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

2,024,785 possible quartets for the 85 sampled specimens representing et al., 2017c). Townsend et al. (2012) proposed theory that allows 80 species. Tetrad uses the algorithm implemented by qQMC to join the the utility of a locus to be quantified by comparing the predicted individual quartet trees into a supertree (Avni et al., 2015). We con- phylogenetic signal supporting a correct resolution (R)versusthe structed a 50% majority-rule consensus tree from 100 nonparametric amount of phylogenetic noise supporting resolution based on bootstrap replicates. homoplasy (H)ofahypotheticalphylogeneticquartet.Wequantified Alternative phylogenetic hypotheses were compared to the max- the difference between R and H to explicitly quantify and visualize imum likelihood tree inferred from the IQ-TREE analysis using the trends of the predicted utility of RADseq data for phylogenetic re- approximately unbiased (AU) test based on the resampling of estimated solution. Specifically, we simultaneously assessed the predicted im- log-likelihoods (RELL) method (Kishino et al., 1990; Shimodaira, 2002). pact of alignment length and temporal depth on RADseq based in- The alternative phylogenetic hypotheses tested include (1) the mono- ferences of short internodes. We quantified R and H for resolving a fi phyly of Nototheniidae as delimited in standard references for sh and short (0.25 million year) internode (t0 in Townsend et al., 2012) notothenioid (DeWitt et al., 1990; Eastman and Eakin, 2000; using increasing quartet depths of 5 or 10 million years beginning 5 Nelson et al., 2016: 465–466), (2) the monophyly of Pleura- million years ago (T in Townsend et al., 2012). For each hypothetical grammatinae, containing Pleuragramma antarctica, Aethotaxis mi- quartet, we increasingly added 500 bp of variable sites from the topteryx, Dissostichus eleginoides, and D. mawsoni (Balushkin, 2000; Near concatenated RADseq alignment to determine changes in R and H as et al., 2007; Near and Cheng, 2008), (3) the monophyly of Lepidono- afunctionofalignmentlength.GiventhatRADseqalignmentsare tothen s.l. (DeWitt et al., 1990), (4) the monophyly of Artedidraco often comprised of over one million variable sites, this approach (Lecointre et al., 2011), (5) the monophyly of Indonotothenia cyano- allows us to assess at what point we would expect the contribution of brancha and sampled species of (DeWitt et al., 1990; R to mitigate any potential impacts of H.Allquantifications were Balushkin, 2000), (6) borchgrevinki as not phylogenetically conducted in the R package PhyInformR (Dornburg et al., 2016)and nested in Trematomus, (7) Trematomus newnesi and T. borchgrevinki are results were visualized as horizon plots in the LatticeExtra Package sister taxa and all other species of Trematomus form a monophyletic (Sarkar, 2008). group (Balushkin, 2000: Fig. 17), which Balushkin classified in Pseu- dotrematomus (Balushkin, 1982), and (8) the phylogeny of Trematomus 3. Results presented in Lautrédou et al. (2012: Fig. 2). The trees with the highest likelihoods consistent with the alternative phylogenies were estimated 3.1. RADseq data using the constraint tree search option in IQ-TREE (Nguyen et al., 2015). We collected RAD data for 80 notothenioid species, representing all major lineages (Supplementary Table 1). After initial quality filtering, 2.3. Quantifying predicted phylogenetic utility we retained an average of 1.7 ×106 reads per sample, subsequently reduced to an average of 52,000 clusters per sample. Each of these We compared RADseq loci to a dataset of legacy markers used in clusters has a minimum sequencing depth of 6×, with an average depth previous investigations of notothenioid relationships with very similar of 23×across all sampled specimens. After additional filtering to re- taxon sampling (Near et al., 2012; Dornburg et al., 2017a) that consisted move clusters with excess heterozygosity, we retained an average of of two mitochondrial genes (16S and ND2), one intron (S7), and five 51,000 consensus sequences per sample. The average estimated het- exons (Rag1, tbr1, SH3PX3, glyt, and zic1; DOI: https://doi.org//10. erozygosity in these clusters is 3.6 ×10−3, with an average estimated 5281/zenodo.801836). We used the program HyPhy (Pond et al., 2005) base calling error rate of 6.0 ×10−4. After clustering consensus se- in the PhyDesign web interface (López-Giráldez and Townsend, 2011) to quences into homologous loci, excluding loci shared by fewer than four quantify site-specific rates of substitution for each legacy marker as well taxa, and filtering identified paralogs, the final dataset contains as the newly generated RADseq data using a publicly available no- 104,709 loci. The average number of loci per specimen is 25,881 tothenioid chronogram (Near et al., 2012; Dornburg et al., 2017a) (SD = 12110). The proportion of missing data in the final concatenated downloaded from zenodo (DOI: https://doi.org//10.5281/zenodo. alignment of all loci is 76.4%. 801836) that was pruned to mirror the taxon sampling of our align- ment. The R package vioplot was used to generate violin plots of rate 3.2. Phylogenomic inference distributions (http://wsopuppenkiste.wiso.uni-goettingen.de/~dadler). By combining a rotated kernel density plot with a box plot of data Maximum likelihood IQ-TREE (MLiq) and SVDquartets analyses of quantiles, violin plots allow for simultaneous visualization of both the the RADseq dataset result in phylogenies that are highly congruent with quartiles and the underlying probability distribution of the site rates one another (Figs. 1 and 2) and very similar to previous inferences (Hintze and Nelson, 1998). Using the equations presented in Townsend using Sanger sequenced legacy markers (Near et al., 2012; Dornburg (2007), phylogenetic informativeness (PI) was quantified for each locus et al., 2017a). The phylogenies were rooted with the three sampled in the R package PhyInformR (Dornburg et al., 2016). Visual detection of species of Bovichtidae and relationships among the non-Antarctic declines in PI over time have been considered a signature of homoplasy lineages and are identical to (Townsend and Leuenberger, 2011) and linked to phylogenetic estima- previous analyses (Near et al., 2012, 2015). These lineages are pruned tion error (Dornburg et al., 2017c). Given that a RADseq dataset consists out of the trees shown in Figs. 1 and 2 to allow focus on the relation- of thousands of loci, PI profiles were visualized using hexagon binning to ships among the Cryonotothenioidea. The MLiq and SVDquartets phy- assess overall trends in PI between loci using the hexbinplot function logenies differ at five nodes: the resolution of Pleuragramma antarctica, package hexbin (http://github.com/edzer/hexbin). The hexbin plots and in four additional apical relationships that are not strongly sup- were compared with PI profiles of the sampled legacy markers and ported in either analysis (Figs. 1 and 2). mapped to the 95% highest probability density interval of notothenioid The MLiq phylogeny resolves two major cryonotothenioid clades, divergence times estimated from previous relaxed molecular clock ana- (1) a clade containing Trematomus, Indonotothenia cyanobrancha, lyses (Near et al., 2012; Dornburg et al., 2017a). Dissostichus, , Pleuragramma antarctica, Although the shape of the PI profiles provides an indication of Lepidonotothen squamifrons, Nototheniops, and and (2) a overall trends of phylogenetic information content, these visualiza- clade containing Gobionotothen, Notothenia, , tions make no explicit quantification of how convergence in char- Artedidraconidae, Bathydraconidae, and Channichthyidae. The acter state not reflecting evolutionary history (noise) will impact Pleuragrammatinae, composed of P. antarctica, A. mitopteryx, and the topological resolution (Townsend and Leuenberger, 2011; Dornburg two species of Dissostichus, is resolved as paraphyletic in both analyses;

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Fig. 1. Maximum likelihood phylogeny of Cryonotothenioidea inferred from RADseq dataset using IQ-TREE. Numbers at nodes are bootstrap values for those with less than 100% support. Photographs of notothenioid specimens by P. Marriott, P. McMillan, R. McPhee, T. J. Near, and C. Struthers and are deposited at the Museum of New Zealand Te Papa Tongarewa and Peabody Museum of Natural History, Yale University.

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Fig. 2. Species tree inferred using SVDquartets. Numbers at nodes are bootstrap support values. Nodes that differ from the maximum likelihood IQ-TREE (Fig. 1) are marked with a filled black circle. Photographs of notothenioid specimens by P. Marriott, P. McMillan, R. McPhee, T. J. Near, and C. Struthers and are deposited at the Museum of New Zealand Te Papa Tongarewa and Peabody Museum of Natural History, Yale University.

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Table 1 Comparison of phylogenetic hypotheses of Notothenioidei using the approximately unbiased (AU) test based on the resampling of estimated log-likelihoods (RELL) method.

Phylogenetic hypothesis logLn ΔlogLn bp-RELL p-AU

Optimal ML tree (Fig. 1) −7765781.407 0.000 0.713 0.774 Nototheniidae monophyletic −7766949.513 1168.106 0.000 0.010 Pleuragrammatinae monophyletic −7765888.625 107.218 0.283 0.250 Lepidonotothen monophyletic −7768175.483 2394.076 0.000 0.007 Artedidraco monophyletic −7767136.315 1354.908 0.000 0.006 Indonotothenia and Notothenia sister taxa −7767942.406 2160.999 0.000 0.001 Pagothenia not nested in Trematomus −7768068.765 2287.358 0.000 0.001 Pseudotrematomus (Balushkin 1982, 2000) −7772505.186 6723.779 0.000 0.001 Phylogeny of Trematomus in Lautredou et al. (2012) −7771618.300 5836.893 0.000 0.001

however, the subclade containing A. mitopteryx and Dissostichus is relationships among species of Trematomus (Fig. 1). Nototheniops, Pa- strongly supported (Fig. 1). The Trematominae, comprising Patagono- tagonotothen, and Lepidonotothen form a clade, but Lepidonotothen as tothen, Nototheniops, L. squamifrons, I. cyanobrancha, and Trematomus, traditionally delimited to contain L. squamifrons and the species of form a clade. Indonotothenia cyanobrancha is resolved as the sister Nototheniops is resolved as non-monophyletic as L. squamifrons is the lineage of Trematomus and there is strong node support for the sister lineage of Patagonotothen (Figs. 1 and 2).

Fig. 3. Predictions of phylogenetic utility for RADseq and legacy DNA sequence datasets. A. Violin plot comparing the distribution of site rates for each dataset, with the size of each da- taset in base pairs (bp) indicated above each plot. An asterisk marks the truncation of the upper tail of the site rate distribution for gra- phical purposes. B. Hexbin plot of the relative phylogenetic informativeness (PI) over time of each RADseq locus. Colors correspond to the number of loci with a measured value of in- formativeness. Curved lines represent PI profiles of each legacy DNA sequence dataset. C. Highest 95% posterior density interval of cryonotothe- nioid divergence times taken from Dornburg et al. (2017a). Dashed vertical lines corre- sponding to the previously estimated most re- cent common ancestor of cryonotothenioids and the onset of rapid lineage diversification hy- pothesized in Near et al. (2012). Photograph of Histiodraco velifer by A. Stewart and is deposited at the Museum of New Zealand Te Papa Ton- garewa.

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Fig. 4. Horizon plots depicting the relationship between sequence length and the predicted prob- ability of phylogenetic noise (H) misleading in- ference based on phylogenetic signal (R). Each row corresponds to a temporal depth of a hypothetical quartet beginning with recent divergences and ex- tending to the K-Pg boundary. Colors indicate the values of R-H, with darker blue colors indicating high R, whites indicating little remaining resolving power; and darker reds indicating strong predicted probabilities of H overwhelming signal. (For inter- pretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Gobionotothen is strongly supported as the sister lineage of an in- geologic time period and estimated 95% highest posterior density in- clusive clade containing Notothenia, Harpagifer, Artedidraconidae, tervals of molecular ages for the origin and diversification in the clade Bathydraconidae, and Channichthyidae (Figs. 1 and 2). The 14 sampled (Fig. 3B & C). Quantification of R, the predicted phylogenetic signal species of Bathydraconidae are a monophyletic group and are resolved supporting a correct resolution, and H, the amount of phylogenetic as sister lineage of the Channichthyidae (Figs. 1 and 2). Relationships noise supporting resolution based on homoplasy, also reveals high within the bathydraconids and channichthyids are well supported in probabilities of R for short internodes through the Late Miocene the MLiq phylogeny, but relationships of the Antarctic Dragonfishes (∼20 Ma; Fig. 4). In all cases, R rapidly maximized with the addition of glacialis, Vomeridens infuscipinnis, and species of Bathydraco data, predicting that RADseq loci contain high levels of phylogenetic are only moderately supported (Fig. 1). Harpagifer and Artedi- information for resolving short internodes. However, our quantifica- draconidae are strongly supported as a clade, but within Artedi- tions reveal that the predicted information content declines at inter- draconidae Artedidraco is paraphyletic with A. skottsbergi resolved as the nodes dating to early periods of the Cenozoic (Fig. 4). By the Cretac- sister lineage of all other artedidraconids and all other sampled species eous-Paleogene boundary (∼66 Ma), there is very little phylogenetic of Artedidraco (A. orianae, A. mirus, A. glareobarbatus, and A. shackletoni) information remaining, even when the dataset contains 106 variable form a clade that is the sister lineage of Dolloidraco longedorsalis (Fig. 1). characters (Fig. 4). Histiodraco velifer and the sampled species of Pogonophryne are resolved as sister lineages (Figs. 1 and 2). 4. Discussion The AU test rejects seven of the eight alternative hypotheses when compared to the optimal tree inferred from the IQ-TREE analysis 4.1. Resolving the phylogenetic history of the notothenioid adaptive (Table 1). The only hypothesis that is not rejected is the monophyly of radiation Pleuragrammatinae (Table 1), which is sampled with Pleuragramma antarctica, Aethotaxis mitopteryx, Dissostichus eleginoides, and D. mawsoni Over the past quarter century, the phylogeny of notothenioids has (Figs. 1 and 2). come increasingly into focus (Balushkin, 1992; Bargelloni et al., 1994, 2000; Lecointre et al., 1997; Balushkin, 2000; Near and Cheng, 2008; 3.3. Predicted phylogenetic information content Dettaï et al., 2012; Near et al., 2012, 2015; Dornburg et al., 2017a); however, there remain considerable areas of uncertainty. The trees A violin plot based comparison of the probability density of calcu- inferred using the RADseq dataset have an unprecedented degree of lated site rates for RADseq loci and other classes of Sanger sequenced resolution and node support for relationships within cryonotothenioid legacy markers reveal that RADseq loci possess a similar distribution of lineages that exhibit high rates of diversification (Fig. 1). The phylo- rates as observed in the previously sampled mtDNA genes (Fig. 3A). geny inferred in this study provides important insight regarding re- Visualizing the density of PI profiles through time for the RADseq loci lationships among species of Trematomus, the phylogeny of species of as a hexagonal heatmap (Fig. 3B), shows a high-predicted level of in- Artedidraconidae and a clarification of the paraphyly of Artedidraco, the formativeness through the Miocene [∼23.0 to 5.3 Ma] (Fig. 3B), with deep paraphyly of Nototheniidae, consistent non-monophyly of Lepi- particularly high densities of informative loci corresponding to the donotothen squamifrons and the species of Nototheniops, the relationships geologic intervals hypothesized as the periods of both the origin and of Bathydraconidae, and continued uncertainty in the phylogenetic diversification of major cryonotothenioid lineages (Fig. 3B & C). Map- placement of Pleuragramma antarctica. The well-supported resolution of ping the PI profiles of all other classes of legacy markers used to in- relationships throughout the notothenioid phylogeny, especially among vestigate notothenioid phylogeny onto this heat map demonstrates little the radiations of closely related species (e.g., Trematomus, Artedi- decline in PI during the interval corresponding to the hypothesized draconidae, Channichthyidae, and Bathydraconidae), provides

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ² opportunities to investigate mechanisms of speciation and evolutionary Pogonophryne (Figs. 1 and 2; Near and Cheng, 2008; Eakin et al., 2009; diversification in one of the most iconic and threatened marine eco- Lecointre et al., 2011; Near et al., 2012); however, the taxonomy of this systems of our planet, an avenue of research identified as one of the six group warrants reevaluation. Species of Pogonophryne are currently priorities for Antarctic science (Kennicutt et al., 2014). classified into five species groups on the basis of external morphological Species of Trematomus comprise an important element of the no- traits (Balushkin and Eakin, 1998). Although previous molecular ana- tothenioid adaptive radiation (Janko et al., 2011; Lautredou et al., lyses support these groupings, the relationships among them remain 2012; Near et al., 2012) as well as a dominant component of the near unresolved (Eakin et al., 2009). Analysis of the RADseq dataset results shore notothenioid fish fauna (Kock, 1992). However, no prior study in a strongly supported phylogenetic hypothesis for the relationships provides strong support for the resolution of the inter-relationships of among the species groups of Pogonophryne; however, the P. mentella these species. The RADseq data reject all of the alternative phylogenetic group, which includes the sampled species P. macropogon, P. cere- relationships involving Trematomus that includes a phylogeny inferred bropogon, P. eakini, and P. fusca, is not monophyletic as P. fusca is the from a set of legacy Sanger sequenced mitochondrial and nuclear genes, sister lineages of P. barsukovi (Figs. 1 and 2). These results suggest a the proposal that Trematomus is limited to T. newnesi, and all other need for a reexamination of the classification of Pogonophryne species species are classified as Pseudotrematomus (Balushkin, 1982), and that diversity using a dataset with greater taxonomic sampling. Given that the , Pagothenia borchgrevinki, is not phylogenetically species of Pogonophryne exhibit substantial overlap in bathymetric nested in Trematomus (Table 1). Previous molecular studies have con- distribution (Eakin, 1990; Eastman, 2017) and diet (Wyanski and sistently resolved T. scotti as the sister lineage to other species of Tre- Targett, 1981; Eakin, 1990; Lombarte et al., 2003), and are character- matomus (Ritchie et al., 1996; Sanchez et al., 2007; Kuhn and Near, ized by very little morphological disparity among the species (Eakin, 2009; Janko et al., 2011; Lautredou et al., 2012; Near et al., 2012), a 1977; Lombarte et al., 2003), increased inter- and intraspecific sam- result also strongly supported in the RADseq inferred phylogenies pling is needed to assess the delimitation of species and investigate the (Figs. 1 and 2). The phylogeny of Trematomus resolves strongly sup- mechanisms driving diversification in this clade. ported monophyletic groups containing epibenthic species T. loenn- In addition to resolving relationships among the most closely related bergii, T. lepidorhinus, and T. eulepidotus as a clade, as well as a clade species, the RADseq inferred phylogenies also provide much-needed containing the demersal species T. bernacchii and T. hansoni (Figs. 1 and resolution to several higher-level relationships within 2). These relationships, initially suggested from analysis of sequence Cryonotothenioidea. Since the first molecular analyses of notothenioid data of mtDNA rRNA genes, are consistent with the expectations of phylogeny, the inclusive family Nototheniidae is consistently resolved habitat use driven adaptive radiation within this clade (Ritchie et al., as a paraphyletic group (Bargelloni et al., 1994; Sanchez et al., 2007; 1996). Dettaï et al., 2012; Near et al., 2012; Dornburg et al., 2017a). The The RADseq phylogenies resolve Indonotothenia cyanobrancha as the concept of Nototheniidae communicated in taxonomic references (e.g., sister lineage of Trematomus (Figs. 1 and 2), a result consistent with Eastman and Eakin, 2000; Nelson et al., 2016) has an origin with the analyses of mtDNA genes and a combination of mtDNA and nuclear pioneering work of Regan (1913: 249–251) and Norman (1938: 7–10) genes (Bargelloni and Lecointre, 1998; Bargelloni et al., 2000; who provided important taxonomic revisions of notothenioid classifi- Dornburg et al., 2017a). Indonotothenia cyanobrancha was long classi- cation. Specifically, Regan (1913) classified species of Harpagifer and fied in the Notothenia (DeWitt et al., 1990), but Balushkin (1984: Artedidraco in Nototheniidae along with Notothenia s.l., Trematomus, 13) described the monotypic genus Indonotothenia to classify this spe- Pleuragramma, Dissostichus, and Eleginops. Norman (1938) classified cies. Based on morphological analyses, Indonotothenia is resolved as the Artedidraco, Dolloidraco, Pogonophryne, and Harpagifer in Harpagifer- sister lineage of a clade containing Notothenia and Paranotothenia idae and limited Nototheniidae to Notothenia s.l., Trematomus, Pleura- (Voskoboinikova, 1993: Fig. 8; Balushkin, 2000: Fig. 15). However, gramma, Dissostichus, and Eleginops. Subsequent discoveries of new assessment of external morphological characters argues for a closer species added Aethotaxis, Cryothenia, and Gvozdarus to Nototheniidae relationship with Trematomus (Hureau, 1970: 225, Table 14). The re- (DeWitt, 1962; Daniels, 1981; Balushkin, 1989). Taxonomic revisions sults of the phylogenetic analyses of the RADseq data strongly reject a by Balushkin (1976, 1992) led to the removal of Eleginops maclovinus close relationship of I. cyanobrancha with other species of Notothenia from Nototheniidae to the monotypic Eleginopsidae and the description and instead suggest that it is a species of Trematomus (Figs. 1 and 2; of several new genera; Gobionotothen, Lepidonotothen, Nototheniops, Table 1). Patagonotothen, and Paranotothenia all of which contain species that Similar to Trematomus, previous molecular phylogenetic analyses do were previously classified as Notothenia (Norman, 1938; Andriashev, not provide strong node support for relationships inferred among spe- 1965). The classification of these disparate lineages as Notothenia at the cies of Artedidraconidae. Artedidraconids occupy benthic habitats of time of the origin of modern classifications for notothenioids con- the Antarctic and exhibit a high diversification rate tributed to the long-held idea that Nototheniidae is a natural group relative to other notothenioid lineages (Eakin, 1990; Near et al., 2012). (e.g., Norman, 1938). Molecular phylogenies, including those inferred Consistent with previous molecular analyses (Derome et al., 2002; from the RADseq data (Figs. 1 and 2), provide a strong inference that Lecointre et al., 2011; Near et al., 2012), the phylogeny inferred from Nototheniidae as traditionally delimited is not monophyletic. The al- the RADseq dataset resolves Artedidraco as paraphyletic, with A. ternative phylogeny with the highest likelihood that depicts Notothe- skottsbergi as the sister lineage of all other artedidraconids (Figs. 1 and niidae, as traditionally delimited, is rejected in the AU test (Table 1). 2). The AU test rejects the best alternative phylogeny that resolves While this study did not sample species of Paranotothenia, previous Artedidraco as a monophyletic group (Table 1). The remaining species molecular phylogenetic analyses consistently resolve Notothenia and of Artedidraco sampled in this study, including the type species A. mirus Paranotothenia as a clade (Cheng et al., 2003; Sanchez et al., 2007; Near (Lönnberg, 1905: 39), comprise a strongly supported clade in the MLiq and Cheng, 2008; Near et al., 2012; Dornburg et al., 2017a). We re- tree that is sister to Dolloidraco longedorsalis (Fig. 1), a relationship not commend that Nototheniidae is limited to Notothenia and Para- hypothesized in previous molecular phylogenetic analyses (Derome notothenia. et al., 2002; Lecointre et al., 2011; Near et al., 2012). The deep para- Similar to the traditional delimitation of Nototheniidae, the RADseq phyly of Artedidraco s.l. and the lack of an available genus group name tree and previous molecular phylogenetic analyses consistently fail to to accommodate A. skottsbergi necessitates a taxonomic revision that resolve Lepidonotothen squamifrons and species of Nototheniops as a reflects the phylogeny and consistent resolution of Artedidraco s.l. as a monophyletic group (Bargelloni et al., 2000; Near and Cheng, 2008; paraphyletic group. Dettaï et al., 2012; Near et al., 2012). Specifically, L. squamifrons is Previous molecular analyses and the RADseq inferred phylogenies resolved as the sister lineage of Patagonotothen (Figs. 1 and 2). In his strongly support monophyly of the species-rich artedidraconid lineage revision of Notothenia, Balushkin (1976, 1979) described the genera

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Lepidonotothen (containing L. squamifrons and the two synonyms L. the two species of Dissostichus (Figs. 1 and 2). Gvozdarus svetovidovi was kempi and L. macrophthalma), Nototheniops (containing N. larseni, and not included in our analysis. Investigation of the uncertainty regarding synonyms N. loesha, N. nybelini, and N. tchizh), and Lindbergichthys phylogenetic relationships of P. antarctica requires additional work as (containing N. mizops, N. nudifrons, and the undescribed N. cf. nudi- evidenced by the inability of the RADseq dataset to reject the best al- frons). The most recent taxonomic revision of these lineages treats ternative phylogeny that depicts Pleuragrammatinae as a monophyletic Nototheniops and Lindbergichthys as subgenera of Lepidonotothen without group (Table 1). identifying any morphological evidence to support the hypothesis that they share common ancestry (DeWitt et al., 1990: 294–295). Phyloge- 4.2. RADseq and Cenozoic radiations netic analysis of morphological characters does not resolve Lepidono- tothen, Nototheniops, and Lindbergichthys as a monophyletic group Rapidly radiating clades characterize some of the most investigated (Balushkin, 2000: Fig. 15) and this phylogeny is rejected in the AU test portions of the Tree of Life (Venditti et al., 2010; Leache et al., 2016; (Table 1). We recommend that Nototheniops (Balushkin, 1976) is the Brennan and Oliver, 2017). Regardless of timescale, the short inter- appropriate genus group name for N. larseni, N. mizops, N. nudifrons, nodes separating species divergences that characterize rapid radiations and N. cf. nudifrons. The monophyly of Nototheniops, as delimited here also present some of the most challenging problems in phylogenetics (Figs. 1 and 2), is supported with several synapomorphic morphological (Sharma et al., 2014; Eytan et al., 2015). As such, assessing the phy- characters that include an upper with perforated scales and logenetic utility of different classes of genomic markers is important for a supraorbital canal with four pores (Andersen, 1984: 24). cost-effective phylogenomic experimental design. Our investigation of The evolutionary loss of red cells, hemoglobin, and the vari- RADseq loci in notothenioids reveals high levels of predicted informa- able loss of expression in Channichthyidae are unique tion for resolving short internodes that date from the Late Oligocene among (Ruud, 1954; Sidell et al., 1997; Sidell and O'Brien, through the Pleistocene (Figs. 3 & 4). This predicted utility is reflected 2006). The genetics and physiological consequences of these highly in high levels of support for most nodes in the notothenioid phylogeny unusual traits are well studied (Egginton and Rankin, 1998; Egginton that correspond to these geologic ages (Fig. 1). However, for radiations et al., 2002; Near et al., 2006; Beers and Sidell, 2009; Beers et al., 2010; dating to the Early Cenozoic (∼65 to 50 mya), our results suggest the Beers and Sidell, 2011; Lewis et al., 2015; Xu et al., 2015; Kuhn et al., massive amount of data offered by RADseq provides diminishing re- 2016); however increased resolution of the phylogenetic relationships turns (Fig. 4). These findings provide quantitative support sub- of Channichthyidae has potential to provide insights into the evolution stantiating claims that RADseq loci are useful for resolving interspecific of hemoglobin loss (Bargelloni et al., 1998; Near et al., 2006; Lau et al., phylogenetic problems (Cariou et al., 2013; Massatti et al., 2016; Eaton 2012). While morphological and molecular phylogenetic analyses et al., 2017), while also setting new expectations of the temporal limits consistently resolve Channichthyidae and Bathydraconidae as a clade for this class of data. Our results suggest RADseq loci contain similar (Iwami, 1985: Fig. 174; Balushkin, 1992: Fig. 6; 2000: Fig. 11; levels of phylogenetic information as the flanking regions of ultra- Bargelloni et al., 2000; Near and Cheng, 2008), most molecular ana- conserved elements (UCEs), which contain high levels of phylogenetic lyses result in phylogenies where the channichthyids are nested in a information for divergences dating to the Miocene (∼20 mya) (Gilbert paraphyletic Bathydraconidae. In these previous molecular analyses et al., 2015). Both the RADseq loci sampled across the diversity of strong support for the resolution of the sister lineage of Chan- notothenioids and UCE-flanking regions in teleost fishes exhibit a rapid nichthyidae is lacking (e.g., Derome et al., 2002; Dettaï et al., 2012; decay in phylogenetic information for divergences arising prior to the Near et al., 2012). The RADseq dataset resolves Bathydraconidae as a K-Pg boundary, 66 mya (Fig. 3; Gilbert et al., 2015). While our study monophyletic group and the sister lineage of Channichthyidae with suggests RADseq loci exhibit high utility for resolving difficult Late strong bootstrap support in the MLiq inferred phylogeny, but with Cenozoic phylogenetic problems, it is important to consider that pre- lower support in the SVDquartet phylogeny (Figs. 1 and 2). The dictions of utility are not guarantees of successful phylogenetic re- common ancestor of Channichthyidae and Bathydraconidae likely ex- solution (Townsend and Leuenberger, 2011). hibited decreased hematocrit, lower concentrations of hemoglobin , and Neither PI profiles nor quartet internode resolution probabilities reduced globin chain multiplicity (D'Avino and Di Prisco, 1988; di make explicit statements on the predicted levels of node support Prisco, 1998; Verde et al., 2007; Wujcik et al., 2007). The phylogenetic (Townsend and Leuenberger, 2011). These approaches merely indicate resolution of the channichthyid sister lineage facilities contextualizing whether there is a probability of phylogenetic information given the the genomic pathways that have given rise to these highly unusual theoretical expectations of phylogenetic experimental design. This does phenotypes. not imply that phylogenetic information sufficient for strong node Despite the well-supported phylogenetic resolution of notothenioid support is present in the dataset. It is possible that limited phylogenetic relationships resulting from analysis of the RADseq dataset, the place- information explains the lack of confidence in the phylogenetic re- ment of the Antarctic Silverfish, Pleuragramma antarctica, remains un- solution of Pleuragramma antarctica and the inability for the RADseq resolved. This monotypic pelagic lineage is a radical departure in dataset to reject the monophyly of the Pleuragrammatinae (Figs. 1 and phenotype from other notothenioids (Eastman, 1997; Voskoboinikova 2; Table 1). Alternatively, the nature of RADseq data imposes several et al., 2017) and is of key importance to Antarctic food webs (Zane challenges to predictions of utility that could also explain our lack of et al., 2006; Mintenbeck and Torres, 2017). Balushkin (2000: S101, confidence in the resolution of these nodes. Fig. 14) delimited the clade Pleuragrammatinae to contain P. antarctica, The low number of variable sites per locus limits the power of ex- the two species of Dissostichus, Aethotaxis mitopteryx, and Gvozdarus perimental design approaches for finely dissecting information content svetovidovi based on pleural ribs originating from the 4th or 5th ver- by locus. While it is possible to generate PI profiles from few or even tebrae and the reduction or absence of a basisphenoid in the skull. single sites, the dependency of the interaction of locus length on R or H Among notothenioids, Pleuragramma antarctica, Dissostichus mawsoni, probabilities renders filtration approaches such as those used in other and Aethotaxis mitopteryx are neutrally buoyant as adults (Eastman and studies not only difficult (Prum et al., 2015; Dornburg et al., 2017b), DeVries, 1982; Near et al., 2003, 2007), and common ancestry of these but potentially misleading. For example, Dornburg et al. (2017b) re- lineages would indicate a single origin of neutral buoyancy in no- cently noted that fast evolving third codon positions in exons captured tothenioids (Near et al., 2007). However, previous molecular analyses by anchored hybrid enrichment are characterized by high instances of do not confidentially support monophyly of Pleuragrammatinae or re- non-random convergence in base frequency (i.e., GC bias). Detecting sult in different resolutions for P. antarctica. Our analyses do not resolve bias using similar approaches is not possible with only a limited number Pleuragrammatinae as a clade, but one poorly supported node separates of variable sites, as convergences that appear slow will not be detected. P. antarctica from a monophyletic group containing A. mitopteryx and In the worst-case scenario, filtering only ‘fast’ sites detected using any

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ² number of methods (e.g., Xia et al., 2003; Goremykin et al., 2009, the Museum of New Zealand Te Papa Tongarewa. This research was 2010) will lead to amplification of an erroneous “signal” in the data supported by the Bingham Oceanographic fund of the Peabody Museum leading to confidence for an erroneous phylogenetic resolution of Natural History, Yale University and the United States National (Dornburg et al., 2017c). Given that we found base frequencies to be Science Foundation ANT-0839007. Two reviewers provided helpful near stationarity (A = 0.265; C = 0.239; G = 0.242; T = 0.253) nu- comments on the manuscript. All data and analysis files have been ar- cleotide bias is not likely a major axis of error in the RADseq dataset. chived on Zenodo (DOI 10.5281/zenodo.1406314). However, even a small number of loci dominated by this or other forms of homoplasy can impact the topological resolution (Shen et al., 2017). Appendix A. Supplementary material Additionally, RADseq datasets do contain high levels of missing data, so it is possible that information rich sites for this node were simply not Supplementary data associated with this article can be found, in the captured during sequencing. As such, determining the factors limiting online version, at https://doi.org/10.1016/j.ympev.2018.09.001. the confident resolution of Pleuragramma antarctica remains an open question. However, in the case of the other inferred notothenioid re- References lationships, congruence in phylogenies inferred from different sets of genes (e.g., Dettaï et al., 2012; Near et al., 2012) and expectations of Abrams, P.A., 2013. How precautionary is the policy governing the Antarctic utility based on analysis of phylogenetic informativeness (Fig. 3) pro- toothfish (Dissostichus mawsoni) fishery? Antarctic Sci. 26, 3–14. fi Andersen, N.C., 1984. Genera and subfamilies of the family Nototheniidae (Pisces, vide con dence in our resolution for the notothenioid phylogeny ) from the Antrarctic and . Steenstrupia 10, 1–34. (Figs. 1 and 2). Andriashev, A.P., 1965. A general review of the Antarctic fish fauna. In: Mieghem, J.V., Understanding the drivers of diversification in rapidly radiating Oye, P.V. (Eds.), Biogeography and Ecology in Antarctica. Junk, The Hague, pp. 491–550. clades is a primary area of research in evolutionary biology (e.g., Avni, E., Cohen, R., Snir, S., 2015. Weighted quartets phylogenetics. Syst. Biol. 64, Gavrilets and Losos, 2009). Given that many of the iconic 233–242. adaptive radiations such as anoles (Poe et al., 2017) and cichlids Balushkin, A.V., 1976. [A short revision of notothenids (Notothenia Richardson and re- (Friedman et al., 2013) are Cenozoic in origin, our findings coupled lated species) from the family Nototheniidae.]. In: Skarlato, A.O., Korovina, V. (Eds.), [Zoogeography and systematics of fish] In Russian, Leningrad, pp. 118–134. with the effectiveness of capturing large amounts of data for non-model Balushkin, A.V., 1979. Lindbergichthys (Nototheniidae) a new generic name for Lindbergia organisms underscore the utility of RADseq for providing phylogenetic Balushkin, 1976 Non Reidel, 1959. J. Ichthyol. 19 (5), 144–145. fi fi resolution to recent radiations and species flocks (e.g., Wagner et al., Balushkin, A.V., 1982. Classi cation of the trematomin shes of Antarctica. In: Kafanov, A.I. (Ed.), Biology of the shelf zones of the World Ocean. USSR Academy of Sciences 2013). However, our study also demonstrates the heterogeneity of Far East Center, Vladivostok. phylogenetic information within this class of genomic markers, offering Balushkin, A.V., 1984. Morphological Bases of the Systematics and Phylogeny of the insights into phylogenetic informativeness as it related to divergence Nototheniid . Academy of Sciences USSR, Zoological Institute, Leningrad. Balushkin, A.V., 1989. Gvozdarus svetovidovi gen. et sp. n. (Pisces, Nototheniidae) from the time (Fig. 3), which is consistent with results presented by Collins and Ross Sea (Antarctic). Zool. Zh. 68, 83–88. Hrbek (2018). While RADseq is of tremendous utility for late Cenozoic Balushkin, A.V., 1992. Classification, phylogenetic relationships, and origins of the fa- radiations, returns increasingly diminish for radiations moving deeper milies of the suborder Notothenioidei (Perciformes). J. Ichthyol. 32 (7), 90–110. Balushkin, A.V., 2000. Morphology, classification, and evolution of notothenioid fishes of in time towards the beginning of the Cenozoic or earlier (Figs. 3 and 4). the Southern Ocean (Notothenioidei, Perciformes). J. Ichthyol. 40, S74–S109. Our results provide an important context for the application of RADseq Balushkin, A.V., Eakin, R.R., 1998. Pogonophryne fusca sp. nov. (Artedidraconidae, data to resolving interspecific phylogenetics and compliment similar Notothenioidei) and notes on the species composition and groups of the genus Pogonophryne Regan. J. Ichthyol. 38, 574–579. studies conducted on other types of next-generation sequence data such Bargelloni, L., Lecointre, G., 1998. Four years in notothenioid systematics: a molecular as UCE or loci captured by anchored hybrid enrichment (Gilbert et al., perspective. In: Prisco, G.D., Pisano, E., Clarke, A. (Eds.), Fishes of Antarctica: A 2015; Prum et al., 2015; Dornburg et al., 2017b; Reddy et al., 2017; biological Overview. Springer-Verlag, Berlin-Heidelberg, pp. 259–273. fi Collins and Hrbek, 2018). Future studies comparing the relative per- Bargelloni, L., Marcato, S., Patarnello, T., 1998. Antarctic sh : evidence for adaptive evolution at subzero temperature. Proc. Nat. Acad. Sci. USA 95, 8670–8675. formance of multiple classes of markers targeted by next-generation Bargelloni, L., Marcato, S., Zane, L., Patarnello, T., 2000. Mitochondrial phylogeny of sequencing techniques will contribute to the optimization of phyloge- notothenioids: a molecular approach to Antarctic fish evolution and biogeography. netic experimental design and lead to an efficient and cost effective Syst. Biol. 49, 114–129. Bargelloni, L., Ritchie, P.A., Patarnello, T., Battaglia, B., Lambert, D.M., Meyer, A., 1994. resolution of the Genomic Tree of Life. Molecular evolution at subzero temperatures: mitochondrial and nuclear phylogenies of fishes from Antarctica (Suborder Notothenioidei), and the evolution of antifreeze Acknowledgements glycopeptides. Mol. Biol. Evol. 11, 854–863. Beers, J.M., Borley, K.A., Sidell, B.D., 2010. Relationship among circulating hemoglobin, nitric oxide synthase activities and angiogenic poise in red- and white-blooded Field and laboratory support was provided by W. Brooks, H. W. Antarctic notothenioid fishes. Comp. Biochem. Phys. A 156, 422–429. Detrich, J. Kendrick, K. L. Kuhn, K.-H. Kock, A. Lamb, J. A. Moore, and Beers, J.M., Sidell, B.D., 2009. Nitric oxide synthase activity correlates with hemoglobin content in Antarctic notothenioid fishes. Integ. Comp. Biol. 49 E198-E198. J. Pennington, G. Watkins-Colwell, K. Zapfe. P. Brickle (South Atlantic Beers, J.M., Sidell, B.D., 2011. Thermal tolerance of Antarctic notothenioid fishes cor- Environmental Research Institute, Falkland Islands), A. L. DeVries and relates with level of circulating hemoglobin. Physiol. Biochem. Zool. 84, 353–362. C.-H. C. Cheng (University of Illinois, USA), D. A. Fernández (Centro Brennan, I.G., Oliver, P.M., 2017. Mass turnover and recovery dynamics of a diverse fi Australian continental radiation. Evolution 71, 1352–1365. Austral de Investigaciones Cientí cas, Argentina), and A. Dettai and G. Cariou, M., Duret, L., Charlat, S., 2013. Is RAD-seq suitable for phylogenetic inference? Lecointre (Muséum National d’Histoire Naturelle, France) provided An in silico assessment and optimization. Ecol. & Evol. 3, 846–852. critical specimens. Fieldwork was facilitated through the United States Cheng, C.-H.C., Chen, L.B., Near, T.J., Jin, Y.M., 2003. Functional antifreeze glycoprotein fi Antarctic Marine Living Resources Program and the officers and crew of genes in temperate-water New Zealand nototheniid sh infer an Antarctic evolu- tionary origin. Mol. Biol. Evol. 20, 1897–1908. the RV Yuzhmorgeologiya, and the 2004 ICEFISH cruise aboard the RVIB Chifman, J., Kubatko, L., 2014. Quartet inference from SNP data under the coalescent Nathaniel B. Palmer. Fresh color images of notothenioids used in Figs. 1 model. Bioinformatics 30, 3317–3324. fi and 2 were taken by P. Marriott, P. McMillan (NIWA), and R. McPhee Clarke, A., Johnston, I.A., 1996. Evolution and adaptive radiation of Antarctic shes. Trends Ecol. Evol. 11, 212–218. (Te Papa). Specimens and data collected by and made available through Collins, R.A., Hrbek, T., 2018. An in silico comparison of protocols for dated phyloge- the New Zealand International Polar Year-Census of Antarctic Marine nomics. Syst. Biol. 67, 633–650. Life Project are gratefully acknowledged. Te Papa collection develop- Colombo, M., Damerau, M., Hanel, R., Salzburger, W., Matschiner, M., 2015. Diversity and disparity through time in the adaptive radiation of Antarctic notothenioid fishes. ment fund AP 2879 “Ross Sea Census of Antarctic Marine Life Inter- J. Evol. Biol. 28, 376–394. national Polar Year (CAML IPY) Voyage”. This work was supported (in Constable, A.J., de la Mare, W.K., Agnew, D.J., Everson, I., Miller, D., 2000. Managing part) by the NZ National Institute of Water and Atmospheric Research fisheries to conserve the Antarctic marine ecosystem: practical implementation of the Convention on the Conservation of Antarctic Marine Living Resources (CCAMLR). Ltd Core Funded Coasts & Oceans Programme 2: Biological Resources ICES J. Mar. Sci. 57, 778–791. subcontract for fundamental knowledge of marine fish biodiversity with

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Cruaud, A., Gautier, M., Galan, M., Foucaud, J., Sauné, L., Genson, G., Dubois, E., Nidelet, noisy data in phylogenomic analyses. J. Mol. Evol. 71, 319–331. S., Deuve, T., Rasplus, J.-Y., 2014. Empirical assessment of RAD sequencing for in- Goremykin, V.V., Viola, R., Hellwig, F.H., 2009. Removal of noisy characters from terspecific phylogeny. Mol. Biol. Evol. 31, 1272–1274. chloroplast genome-scale data suggests revision of phylogenetic placements of D'Avino, R., Di Prisco, G., 1988. Antarctic fish hemoglobin: an outline of the molecular Amborella and Ceratophyllum. J. Mol. Evol. 68, 197–204. structure and binding properties—I. Molecular structure. Comparative Herrera, S., Shank, T.M., 2016. RAD sequencing enables unprecedented phylogenetic Biochem. Physiol. Part B: Comparative Biochem. 90, 579–584. resolution and objective species delimitation in recalcitrant divergent taxa. Mol. Daniels, R.A., 1981. Cryothenia peninsulae, a new genus and species of notothenioid fish Phylogenet. Evol. 100, 70–79. from the Antarctic Peninsula. Copeia 1981, 559–562. Hintze, J.L., Nelson, R.D., 1998. Violin plots: a box plot-density trace synergism. Am. Derome, N., Chen, W.-J., Dettai, A., Bonillo, C., Lecointre, G., 2002. Phylogeny of Statist. 52, 181–184. Antarctic dragonfishes (Bathydraconidae, Notothenioidei, Teleostei) and related fa- Hoang, D.T., Chernomor, O., von Haeseler, A., Minh, B.Q., Vinh, L.S., 2018. UFBoot2: milies based on their anatomy and two mitochondrial genes. Mol. Phylogenet. Evol. improving the ultrafast bootstrap approximation. Mol. Biol. Evol. 35, 518–522. 24, 139–152. Hureau, J.-C., 1970. Biologie comparee de quelques Poissons antarctiques Dettaï, A., Berkani, M., Lautredou, A.C., Couloux, A., Lecointre, G., Ozouf-Costaz, C., (Nototheniidae). Bull. Inst. oceanogr. Monaco 1–244. Gallut, C., 2012. Tracking the elusive monophyly of nototheniid fishes (Teleostei) Ingram, T., Mahler, D.L., 2011. Niche diversification follows key innovation in Antarctic with multiple mitochondrial and nuclear markers. Mar. Genom. 8, 49–58. fish radiation. Mol. Ecol. 20, 4590–4591. DeWitt, H.H., 1962. A new Antarctic nototheniid fish with notes on two recently de- Iwami, T., 1985. Osteology and relationships of the family Channichthyidae. Mem. Nat. scribed nototheniiforms. Copeia 1962, 826–833. Inst. Polar Res. Tokyo 36, 1–69. DeWitt, H.H., Heemstra, P.C., Gon, O., 1990. Nototheniidae. In: Gon, O., Heemstra, P.C. Janko, K., Marshall, C., Musilova, Z., Van Houdt, J., Couloux, A., Cruaud, C., Lecointre, (Eds.), Fishes of the Southern Ocean. J.L.B. Smith Institute of Ichthyology, G., 2011. Multilocus analyses of an Antarctic fish species flock (Teleostei, Grahamstown, South Africa, pp. 279–331. Notothenioidei, Trematominae): phylogenetic approach and test of the early-radia- di Prisco, G., 1998. Molecular adaptations of Antarctic fish hemoglobins. In: Prisco, G.D., tion event. Mol. Phylogenet. Evol. 60, 305–316. Pisano, E., Clarke, A. (Eds.), Fishes of Antarctica: A Biological Overview. Springer- Kalyaanamoorthy, S., Minh, B.Q., Wong, T.K.F., von Haeseler, A., Jermiin, L.S., 2017. Verlag, pp. 339–353. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods Dornburg, A., Federman, S., Lamb, A.D., Jones, C.D., Near, T.J., 2017a. Cradles and 14, 587–589. museums of Antarctic teleost biodiversity. Nature Ecol. Evol. 1, 1379–1384. Kennicutt, M.C., Chown, S.L., Cassano, J.J., Liggett, D., Massom, R., Peck, L.S., Rintoul, Dornburg, A., Fisk, J.N., Tamagnan, J., Townsend, J.P., 2016. PhyInformR: phylogenetic S.R., Storey, J.W.V., Vaughan, D.G., Wilson, T.J., Sutherland, W.J., 2014. Polar re- experimental design and phylogenomic data exploration in R. BMC Evol. Biol. 16, search: six priorities for Antarctic science. Nature 512, 23–25. 262. Kishino, H., Miyata, T., Hasegawa, M., 1990. Maximum likelihood inference of protein Dornburg, A., Su, Z., Townsend, J.P., 2018. Optimal rates for phylogenetic inference and phylogeny and the origin of chloroplasts. J. Mol. Evol. 31, 151–160. experimental design in the era of genome-scale datasets. Syst. Biol syy047-syy047. Kock, K.H., 1992. Antarctic and Fisheries. Cambridge University Press, Cambridge. Dornburg, A., Townsend, J.P., Brooks, W., Spriggs, E., Eytan, R.I., Moore, J.A., Kuhn, D.E., O'Brien, K.M., Crockett, E.L., 2016. Expansion of capacities for iron transport Wainwright, P.C., Lemmon, A.R., Lemmon, E.M., Near, T.J., 2017b. New insights on and sequestration reflects plasma volumes and heart mass among white-blooded the sister lineage of percomorph fishes with an anchored hybrid enrichment dataset. notothenioid fishes. Am. J. Physiol.-Reg. I 311, R649–R657. Mol. Phylogenet. Evol. 110, 27–38. Kuhn, K.L., Near, T.J., 2009. Phylogeny of Trematomus (Notothenioidei: Nototheniidae) Dornburg, A., Townsend, J.P., Wang, Z., 2017c. Maximizing power in phylogenetics and inferred from mitochondrial and nuclear gene sequences. Antarctic Sci. 21, 565–570. phylogenomics: a perspective illuminated by fungal big data. In: Townsend, J.P., La Mesa, M., Eastman, J.T., Vacchi, M., 2004. The role of notothenioid fish in the food Wang, Z. (Eds.), Advances in Genetics. Academic Press, pp. 1–47. web of the Ross Sea shelf waters: a review. Polar Biol. 27, 321–338. Eakin, R., 1990. Artedidraconidae. In: Gon, O., Heemstra, P.C. (Eds.), Fishes of the Lau, Y.T., Parker, S.K., Near, T.J., Detrich, H.W., 2012. Evolution and function of the Southern Ocean. J.L.B. Smith Institute of Ichthyology, Grahamstown, South Africa, globin intergenic regulatory regions of the Antarctic dragonfishes (Notothenioidei: pp. 332–356. Bathydraconidae). Mol. Biol. Evol. 29, 1071–1080. Eakin, R.R., 1977. Morphology and distribution of species in the genus Pogonophryne Lautredou, A.C., Hinsinger, D.D., Gallut, C., Cheng, C.H.C., Berkani, M., Ozouf-Costaz, C., (Pisces, Harpagiferidae). In: Pawson, D.L., Kornicker, L.S. (Eds.), Antarcic Research Cruaud, C., Lecointre, G., Dettai, A., 2012. Phylogenetic footprints of an Antarctic Series, vol. 28, Biology of the Antarctic Seas VIII. American Geophysical Union, radiation: The Trematominae (Notothenioidei, Teleostei). Mol. Phylogenet. Evol. 65, Washington, pp. 1–20. 87–101. Eakin, R.R., Eastman, J.T., Near, T.J., 2009. A new species and a molecular phylogenetic Leache, A.D., Banbury, B.L., Linkem, C.W., de Oca, A.N., 2016. Phylogenomics of a rapid analysis of the Antarctic fish genus Pogonophryne (Notothenioidei: Artedidraconidae). radiation: is chromosomal evolution linked to increased diversification in North Copeia 2009, 705–713. American spiny lizards (Genus Sceloporus)? BMC Evol. Biol. 16, 63. Eastman, J.T., 1993. Antarctic Fish Biology: Evolution in a Unique Environment. Lecointre, G., Bonillo, C., Ozouf-Costaz, C., Hureau, J.-C., 1997. Molecular evidence for Academic Press, San Diego. the origins of Antarctic fishes: paraphyly of the Bovichtidae and no indication for the Eastman, J.T., 1997. Phyletic divergence and specialization for pelagic life in the monophyly of the Notothenioidei (Teleostei). Polar Biol. 18, 193–208. Antarctic nototheniid fish Pleuragramma antarcticum. Comp. Biochem. Phys. 118A, Lecointre, G., Gallut, C., Bonillo, C., Couloux, A., Ozouf-Costaz, C., Dettai, A., 2011. The 1095–1101. Antarctic fish genus Artedidraco is paraphyletic (Teleostei, Notothenioidei, Eastman, J.T., 2005. The nature of the diversity of Antarctic fishes. Polar Biol. 28, Artedidraconidae). Polar Biol. 34, 1135–1145. 93–107. Lewis, J.M., Grove, T.J., O'Brien, K.M., 2015. Energetic costs of protein synthesis do not Eastman, J.T., 2017. Bathymetric distributions of notothenioid fishes. Polar Biol. 40, differ between red- and white-blooded Antarctic notothenioid fishes. Comp. Biochem. 2077–2095. Phys. A 187, 177–183. Eastman, J.T., DeVries, A.L., 1982. Buoyancy studies of notothenioid fishes in McMurdo Lombarte, A., Olaso, I., Bozzano, A., 2003. Ecomorphological trends in the Sound, Antarctica. Copeia 1982, 385–393. Artedidraconidae (Pisces: Perciformes: Notothenioidei) of the Weddell Sea. Antarctic Eastman, J.T., Eakin, R.R., 2000. An updated species list of notothenioid fish Sci. 15, 211–218. (Perciformes; Notothenioidei), with comments on Antarctic species. Arch. Fish. Mar. Lönnberg, E., 1905. The fishes of the Sweedish South Polar Expedition. Wissenschaftliche Res. 48, 11–20. Ergebnisse der Schwedischen Südpolar-Expedition 1901–1903 (5), 1–69. Eaton, D.A.R., 2014. PyRAD: assembly of de novo RADseq loci for phylogenetic analyses. López-Giráldez, F., Townsend, J.P., 2011. PhyDesign: an online application for profiling Bioinformatics 30, 1844–1849. phylogenetic informativeness. BMC Evol. Biol. 11. Eaton, D.A.R., Spriggs, E.L., Park, B., Donoghue, M.J., 2017. Misconceptions on missing Massatti, R., Reznicek, A.A., Knowles, L.L., 2016. Utilizing RADseq data for phylogenetic data in RAD-seq phylogenetics with a deep-scale example from flowering plants. Syst. analysis of challenging taxonomic groups: a case study in Carex sect. Racemosae. Am. Biol. 66, 399–412. J. Bot. 103, 337–347. Egginton, S., Rankin, J.C., 1998. Vascular adaptations for a low pressure/high flow blood Matschiner, M., Hanel, R., Salzburger, W., 2011. On the origin and trigger of the no- supply to locomotory muscles of Antarctic icefish. In: Di Prisco, G., Pisano, E., Clarke, tothenioid adaptive radiation. Plos One 6, e18911. A. (Eds.), Fishes of Antarctica: A Biological Overview. Springer Milan, Milano, pp. Mintenbeck, K., Torres, J.J., 2017. Impact of climate change on the Antarctic Silverfish 185–195. and its consequences for the Antarctic ecosystem. In: Vacchi, M., Pisano, E., Egginton, S., Skilbeck, C., Hoofd, L., Calvo, J., Johnston, I.A., 2002. Peripheral oxygen Ghigliotti, L. (Eds.), The Antarctic Silverfish: A Keystone Species in a Changing transport in skeletal muscle of Antarctic and sub-Antarctic notothenioid fish. J. Exp. Ecosystem. Springer International Publishing, Cham, pp. 253–286. Biol. 205, 769–779. Near, T.J., Cheng, C.-H.C., 2008. Phylogenetics of notothenioid fishes (Teleostei: Eytan, R.I., Evans, B.R., Dornburg, A., Lemmon, A.R., Lemmon, E.M., Wainwright, P.C., Acanthomorpha): inferences from mitochondrial and nuclear gene sequences. Mol. Near, T.J., 2015. Are 100 enough? Inferring acanthomorph teleost phylogeny using Phylogenet. Evol. 47, 832–840. Anchored Hybrid Enrichment. BMC Evol. Biol. 15. Near, T.J., Dornburg, A., Harrington, R.C., Oliveira, C., Pietsch, T.W., Thacker, C.E., Friedman, M., Keck, B.P., Dornburg, A., Eytan, R.I., Martin, C.H., Hulsey, C.D., Satoh, T.P., Katayama, E., Wainwright, P.C., Eastman, J.T., Beaulieu, J.M., 2015. Wainwright, P.C., Near, T.J., 2013. Molecular and fossil evidence place the origin of Identification of the notothenioid sister lineage illuminates the biogeographic history cichlid fishes long after Gondwanan rifting. Proc. R. Soc. B 280, 20131733. of an Antarctic adaptive radiation. BMC Evol. Biol. 15, 109. Gavrilets, S., Losos, J.B., 2009. Adaptive radiation: contrasting theory with data. Science Near, T.J., Dornburg, A., Kuhn, K.L., Eastman, J.T., Pennington, J.N., Patarnello, T., Zane, 323, 732–737. L., Fernandez, D.A., Jones, C.D., 2012. Ancient climate change, antifreeze, and the Gilbert, P.S., Chang, J., Pan, C., Sobel, E.M., Sinsheimer, J.S., Faircloth, B.C., Alfaro, M.E., evolutionary diversification of Antarctic fishes. Proc. Nat. Acad. Sci. USA 109, 2015. Genome-wide ultraconserved elements exhibit higher phylogenetic informa- 3434–3439. tiveness than traditional gene markers in percomorph fishes. Mol. Phylogenet. Evol. Near, T.J., Kendrick, B.J., Detrich, H.W., Jones, C.D., 2007. Confirmation of neutral 92, 140–146. buoyancy in Aethotaxis mitopteryx DeWitt (Notothenioidei: Nototheniidae). Polar Goremykin, V.V., Nikiforova, S.V., Bininda-Emonds, O.R.P., 2010. Automated removal of Biol. 30, 443–447.

 T.J. Near et al. 0ROHFXODU3K\ORJHQHWLFVDQG(YROXWLRQ  ²

Near, T.J., Parker, S.K., Detrich, H.W., 2006. A genomic fossil reveals key steps in he- 1791–1802. moglobin loss by the Antarctic icefishes. Mol. Biol. Evol. 23, 2008–2016. Sidell, B.D., Vayda, M.E., Small, D.J., Moylan, T.J., Londraville, R.L., Yuan, M.L., Rodnick, Near, T.J., Russo, S.E., Jones, C.D., DeVries, A.L., 2003. Ontogenetic shift in buoyancy K.J., Eppley, Z.A., Costello, L., 1997. Variable expression of myoglobin among the and habitat in the Antarctic toothfish, Dissostichus mawsoni (Perciformes: hemoglobinless Antarctic icefishes. Proc. Nat. Acad. Sci. USA 94, 3420–3424. Nototheniidae). Polar Biol. 26, 124–128. Steel, M., Leuenberger, C., 2017. The optimal rate for resolving a near-polytomy in a Nelson, J.S., 1994. Fishes of the World, third ed. Wiley, New York. phylogeny. J. Theoret. Biol. 420, 174–179. Nelson, J.S., Grande, T.C., Wilson, M.V.H., 2016. Fishes of the World. John Wiley & Sons Su, Z., Wang, Z., López-Giráldez, F., Townsend, J.P., 2014. The impact of incorporating Inc, Hoboken. molecular evolutionary model into predictions of phylogenetic signal and noise. Nguyen, L.T., Schmidt, H.A., von Haeseler, A., Minh, B.Q., 2015. IQ-TREE: a fast and Front. Ecol. Evol. 2, 11. effective stochastic algorithm for estimating maximum likelihood phylogenies. Mol. Townsend, J.P., 2007. Profiling phylogenetic informativeness. Syst. Biol. 56, 222–231. Biol. Evol. 32, 268–274. Townsend, J.P., Leuenberger, C., 2011. Taxon sampling and the optimal rates of evolution Norman, J.R., 1938. Coast fishes. Part III. The Antarctic zone. Discorery Rep. 18, 1–104. for phylogenetic inference. Syst. Biol. 60, 358–365. Poe, S., Nieto-Montes de Oca, A., Torres-Carvajal, O., De Queiroz, K., Velasco, J.A., Townsend, J.P., Su, Z., Tekle, Y.I., 2012. Phylogenetic signal and noise: predicting the Truett, B., Gray, L.N., Ryan, M.J., Kohler, G., Ayala-Varela, F., Latella, I., 2017. A power of a data set to resolve phylogeny. Syst. Biol. 61, 835–849. phylogenetic, biogeographic, and taxonomic study of all extant species of Anolis Venditti, C., Meade, A., Pagel, M., 2010. Phylogenies reveal new interpretation of spe- (Squamata; Iguanidae). Syst. Biol. 66, 663–697. ciation and the Red Queen. Nature 463, 349–352. Pond, S.L.K., Frost, S.D.W., Muse, S.V., 2005. HyPhy: hypothesis testing using phylo- Verde, C., Lecointre, G., di Prisco, G., 2007. The phylogeny of polar fishes and the genies. Bioinformatics 21, 676–679. structure, function and molecular evolution of hemoglobin. Polar Biol. 30, 523–539. Prum, R.O., Berv, J.S., Dornburg, A., Field, D.J., Townsend, J.P., Lemmon, E.M., Lemmon, Voskoboinikova, O., Detrich III, H.W., Albertson, R.C., Postlethwait, J.H., Ghigliotti, L., A.R., 2015. A comprehensive phylogeny of birds (Aves) using targeted next-genera- Pisano, E., 2017. Evolution reshaped life for the water column: the skeleton of the tion DNA sequencing. Nature 526, 569–573. Antarctic Silverfish Pleuragramma antarctica (Boulenger 1902). In: Vacchi, M., Reddy, S., Kimball, R.T., Pandey, A., Hosner, P.A., Braun, M.J., Hackett, S.J., Han, K.L., Pisano, E., Ghigliotti, L. (Eds.), The Antarctic Silverfish. A Keystone Species in a Harshman, J., Huddleston, C.J., Kingston, S., Marks, B.D., Miglia, K.J., Moore, W.S., Changing Ecosystem. Springer International Publishing, Cham, pp. 3–26. Sheldon, F.H., Witt, C.C., Yuri, T., Braun, E.L., 2017. Why do phylogenomic data sets Voskoboinikova, O.S., 1993. Evolution of the visceral skeleton and phylogeny of the yield conflicting trees? Data type influences the Avian Tree of Life more than taxon Nototheniidae. J. Ichthyol. 33, 23–47. sampling. Syst. Biol. 66, 857–879. Wagner, C.E., Keller, I., Wittwer, S., Selz, O.M., Mwaiko, S., Greuter, L., Sivasundar, A., Regan, C.T., 1913. The antarctic fishes of the Scottish national antarctic expedition. Seehausen, O., 2013. Genome-wide RAD sequence data provide unprecedented re- Trans. R. Soc. Edinburgh 49, 229–292. solution of species boundaries and relationships in the Lake Victoria cichlid adaptive Ritchie, P.A., Bargelloni, L., Meyer, A., Taylor, J.A., Macdonald, J.A., Lambert, D.M., radiation. Mol. Ecol. 22, 787–798. 1996. Mitochondrial phylogeny of trematomid fishes (Nototheniidae, Perciformes) Wang, X., Ye, X., Zhao, L., Li, D., Guo, Z., Zhuang, H., 2017. Genome-wide RAD se- and the evolution of Antarctic fish. Mol. Phylogenet. Evol. 5, 383–390. quencing data provide unprecedented resolution of the phylogeny of temperate Ruud, J.T., 1954. Vertebrates without erythrocytes and blood pigment. Nature 173, bamboos (Poaceae: Bambusoideae). Sci. Rep. 7, 11546. 848–850. Whitfield, J.B., Lockhart, P.J., 2007. Deciphering ancient rapid radiations. Trends Ecol. Sanchez, S., Dettai, A., Bonillo, C., Ozouf-Costaz, C., Detrich, H.W., Lecointre, G., 2007. Evol. 22, 258–265. Molecular and morphological phylogenies of the Antarctic teleostean family Wujcik, J.M., Wang, G., Eastman, J.T., Sidell, B.D., 2007. Morphometry of retinal vas- Nototheniidae, with emphasis on the Trematominae. Polar Biol. 30, 155–166. culature in Antarctic fishes is dependent upon the level of hemoglobin in circulation. Sarkar, D., 2008. Lattice: Multivariate Data Visualization with R. Springer-Verlag, New J. Exp. Biol. 210, 815–824. York. Wyanski, D.M., Targett, T.E., 1981. Feeding biology of fishes in the endemic Antarctic Sharma, P.P., Kaluziak, S.T., Perez-Porro, A.R., Gonzalez, V.L., Hormiga, G., Wheeler, Harpagiferidae. Copeia 1981, 686–693. W.C., Giribet, G., 2014. Phylogenomic interrogation of arachnida reveals systemic Xia, X., Xie, Z., Salemi, M., Chen, L., Wang, Y., 2003. An index of substitution saturation conflicts in phylogenetic signal. Mol. Biol. Evol. 31, 2963–2984. and its application. Mol. Phylogenet. Evol. 26, 1–7. Shen, X.-X., Hittinger, C.T., Rokas, A., 2017. Contentious relationships in phylogenomic Xu, Q., Cai, C., Hu, X., Liu, Y., Guo, Y., Hu, P., Chen, Z., Peng, S., Zhang, D., Jiang, S., Wu, studies can be driven by a handful of genes. Nature Ecol. Evol. 1, 0126. Z., Chan, J., Chen, L., 2015. Evolutionary suppression of erythropoiesis via the fi Shimodaira, H., 2002. An approximately unbiased test of phylogenetic tree selection. modulation of TGF-β signalling in an Antarctic ice sh. Mol. Ecol. 24, 4664–4678. Syst. Biol. 51, 492–508. Zane, L., Marcato, S., Bargelloni, L., Bortolotto, E., Papetti, C., Simonato, M., Varotto, V., Sidell, B.D., O'Brien, K.M., 2006. When bad things happen to good fish: the loss of he- Patarnello, T., 2006. Demographic history and population structure of the Antarctic moglobin and myoglobin expression in Antarctic icefishes. J. Exp. Biol. 209, silverfish Pleuragramma antarcticum. Mol. Ecol. 15, 4499–4511.